| Abstract |
Silica scale problem at geothermal power plants is one of the biggest issues to be solved to promote much install capacity because of decrease in power generation efficiency. However, the formation mechanism of silica scale has not yet been clarified. Previous geochemical investigations show that Al and Fe in geothermal water are closely associated with the promotion of silica scale formation. However, no study has been done to quantitatively estimate the contribution factors of each geochemical parameter to assess the countermeasure. Therefore, the purpose of this study is to create a machine learning model based on geochemical data and conduct factor analysis using partial derivative method. As for the geochemical data, the input data is geothermal water properties, and the output data is the weight of silica scale. Preliminary results showed that the time has positive contribution (46.0%), and the kinetics of monosilisic acid and reactive aluminum are also positive contribution (SiMKinetic_05_10: 5.7%, SiMKinetic_20_30: 3.3 %.). Though some of the Al parameters negatively contributes to weight of silica scale (AlRKinetic_45_60 : 4.0 %, AlR : 2.1%, Al-slope: 1.0%), there is no contribution of Fe parameters. The results from machine learning indicated that the silica scale weight is promoted with time past, and the Kinetic parameter is significant to predict. In addition, the chemical state of responsible species such as reactive Al may play an important role to inhibit the formation of silica scale as well based on the machine learning. |